Modeling Unemployment Rates by Race and Gender: A Nonlinear Time Series Approach

This paper presents an unemployment rate model that provides insight into how the time series behavior, in terms of both the mean and volatility, of the unemployment rates of black males, white males, black females, and white females differ. Demographic differences in the unemployment rate response are likely to occur if certain demographic groups face discrimination or if different demographic groups gave differing investments in human capital, for example. In addition, there may be differences in other characteristics of the groups, such as differences in the age of distribution or in the marital status distribution. This paper develops and estimates a model to determine whether or not differences in unemployment rate volatility among demographic groups actually exist, utilizing an ARCH-class (autoregressive conditional heteroscedasticity) model. The findings suggest that conditional variance is symmetric for white females, black females, and black males, but is asymmetric for white males. In particular, the findings indicate that innovations increase the conditional volatility changes in each group's unemployment rate and have symmetric effects for all groups except white males.